Remove decodeURIComponent method for KVv2 secret path on list view#28698
Remove decodeURIComponent method for KVv2 secret path on list view#28698Monkeychip merged 14 commits intomainfrom
Conversation
github-actions
bot
added
the
hashicorp-contributed-pr
|
CI Results: |
Monkeychip
changed the title
Monkeychip
changed the title
|
Build Results: |
hellobontempo
left a comment
hellobontempo
left a comment
There was a problem hiding this comment.
Thanks for addressing this and adding test coverage! Just comments about the subtext, as it's not quite accurate right now. Looks good otherwise
| assert | ||
| .dom(PAGE.breadcrumbAtIdx(3)) | ||
| .hasText('foo%2fbar', 'foo%2fbar is the second directory and shows up as a separate breadcrumb'); | ||
| assert.dom(PAGE.breadcrumbAtValue('world')).exists('the current breadcrumb is value world'); |
There was a problem hiding this comment.
How do we know from this assertion that it's the current breadcrumb? It seems like this might be better as
| assert.dom(PAGE.breadcrumbAtValue('world')).exists('the current breadcrumb is value world'); | |
| assert.dom(PAGE.breadcrumbAtIdx(4)).hasText('world', 'the current breadcrumb is value world'); |
There was a problem hiding this comment.
That was my original path, but it can't find the Idx4 because it's not a link but the active page view. I could do an additional check for class hds-breadcrumb__current.
or I could create a new selector that does not include the a. Below is the current selector:
[data-test-breadcrumbs] li:nth-child(${idx + 1}) a
There was a problem hiding this comment.
Gotcha, that makes sense. Using classes isn't the greatest because they can change, but I'm in favor of that, since it's clearer that's what we're testing. Then we don't have to add another selector.
That, or if we could update the existing breadcrumb selector - which doesn't look like it's used much or at all to take a value arg
There was a problem hiding this comment.
Removed the breadcrumb selector and it borked things even though I too couldn't find a usage.
Ended up creating a breadcrumbCurrentAtIdx which I feels like is valuable. It tells us we're confirming location and that it's a current status.
ui/tests/acceptance/secrets/backend/kv/kv-v2-workflow-navigation-test.js
Outdated
Show resolved
Hide resolved
ui/tests/acceptance/secrets/backend/kv/kv-v2-workflow-navigation-test.js
Outdated
Show resolved
Hide resolved
…28698) * remove encoding for KVv2 * test coverage * changelog * validations * Revert "validations" This reverts commit d6fd291. * update subtext for secret path * Update list.js * Update secret-edit.js * test coverage for data-octets * Update list-directory.js * fix modelForm test * amend subText * test selector things
* Remove decodeURIComponent method for KVv2 secret path on list view (#28698) * remove encoding for KVv2 * test coverage * changelog * validations * Revert "validations" This reverts commit d6fd291. * update subtext for secret path * Update list.js * Update secret-edit.js * test coverage for data-octets * Update list-directory.js * fix modelForm test * amend subText * test selector things * fix / update to old test selectors
2 participants
Description
Fixes Issue #25905.
Previously, the KVv2 secretPath was run through a
decodeURIComponentmethod while rendering the list view. This method found percent encoded data-octets and decoded them. In practice this means that something like%2fwould decode to/. This is problematic for a user not intending to make a directory during path name creation but include the%2fas part of the path name.This fix only applies to KVv2 because KVv1, Cubbyhole, etc pull the path name from the url where the URL is already encoded so decoding something like
hello%252fmeepwould equalhello%2fmeep. Note: there is another bug I where these engines cannot render the create page on initial transitions for paths that contain data-octets. I've ticketed this bug and will address in a separate PR.Video of before fix and after fix
Screen.Recording.2024-10-15.at.9.07.25.AM.mov